Hugging Face Daily PapersTan-Dzung Do, Tuan Dat Phuong, Nico BohlingerSep 29, 20261 min readpaperadvanced

CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments

Summary

CrossBFM distills a shared latent behavior space for humanoids, enabling a single vector to represent motion, pose, or reward across different robot embodiments. It drastically cuts training time from hundreds to ~11 GPU-hours and allows cross-embodiment transfer and generalization to unseen robots.

CrossBFM creates a unified latent behavior space transferable across various humanoid embodiments.
Training time is reduced from hundreds of GPU-hours per robot to ~11 GPU-hours for multiple robots.
The unified encoder architecture uses no robot-specific parameters, enabling simultaneous training.
It supports motion tracking, goal reaching, and reward optimization across distilled humanoids.

This work is crucial for robotics engineers and researchers, as it accelerates the development of generalizable humanoid control policies and reduces computational costs for training.

8/10

machine learning reinforcement learning ai robotics deep learning

Related reading

Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation arxiv.org
PRISM is a real-to-sim-to-real framework that uses counterfactual video generation to create diverse training data for humanoid loco-manipulation. It enables training a single policy that generalizes to unseen objects without real-world fine-tuning, addressing data scarcity in robotics.
Hugging Face Daily Papersarxiv.orgZihan Wang, Zhen Wu, Pieter Abbeel7d1 minpaper
Grounded Action Model: 3D Grounding as a Foundation for Robotics arxiv.org
The Grounded Action Model (GAM) adds explicit 3D metric grounding to robot foundation models via a shared object‑centric representation, improving robustness to scene changes. Experiments on RoboTwin 2.0, LIBERO‑PRO, and real robots show state‑of‑the‑art success rates, especially under visual shift and long‑horizon tasks.
Hugging Face Daily Papersarxiv.orgGehao Zhang, Weikai Huang, Shailesh Shailesh16d1 minpaper
Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models arxiv.org
Latent Interface Training (LIT) first learns a goal‑conditioned action prior without visual input, then adds a pose‑supervised latent interface as the only visual conditioning path. Applied to several vision‑language‑action models, LIT cuts vision‑action shortcuts and lifts LIBERO‑Plus success by 3.9–10.7 points and real‑world task success by 13.3–16.7 points under distribution shifts.
Hugging Face Daily Papersarxiv.orgJianman Lin, Shailesh Shailesh, Zhongyi Luo25d1 minpaper
EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Tasks arxiv.org
EMem‑Bench is a new benchmark of 2,554 long‑horizon embodied episodes that explicitly tests an agent’s ability to construct, update, and reuse memory across four defined challenges. The authors also release EMem, a spatial‑event‑scene external memory, and an 8B policy (EMem‑8B) that together achieve the strongest performance, highlighting persistent gaps in current multimodal LLM agents.
Hugging Face Daily Papersarxiv.orgLizhou Liang, Xinyu Zhong, Miao Pan13d1 minpaper
HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness arxiv.org
HarnessVLN introduces a zero‑shot, training‑free embodied navigation framework that wraps a multimodal LLM in an "Agent Harness" – a tool‑based protocol that validates planner actions against spatial evidence, tracks progress with hierarchical event memory, and maintains a persistent spatiotemporal graph for recovery. The system works for instruction‑following and object‑goal tasks, achieving 60.…
Hugging Face Daily Papersarxiv.orgYang Chen, Lirong Che, Zhenyu Huang22d1 minpaper
Dream4ACT: A Shared Visual Action Interface for Multi-Embodiment Video-Action Modeling arxiv.org
Dream4ACT introduces a shared visual action interface (“action views”) that renders robot joint configurations from four virtual cameras using URDF forward kinematics, letting a single video autoencoder and diffusion transformer model both observations and actions across different robot embodiments. Trained with masked flow‑matching, the model attains 88.98% success on RoboTwin 2.0 and a 65.66 ov…
Hugging Face Daily Papersarxiv.orgXiangyu Zhu, Jin Xu, Yue Guo6d1 minpaper

CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments

Summary

Related reading

Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulationarxiv.org

Grounded Action Model: 3D Grounding as a Foundation for Roboticsarxiv.org

Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Modelsarxiv.org

EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Tasksarxiv.org

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harnessarxiv.org

Dream4ACT: A Shared Visual Action Interface for Multi-Embodiment Video-Action Modelingarxiv.org

Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation arxiv.org

Grounded Action Model: 3D Grounding as a Foundation for Robotics arxiv.org

Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models arxiv.org

EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Tasks arxiv.org

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness arxiv.org

Dream4ACT: A Shared Visual Action Interface for Multi-Embodiment Video-Action Modeling arxiv.org